Re: [SAGE] number of eggs in a basket

["Bevan C. Bennett" <bevan@fulcrummicro.com>]

Jan Schaumann wrote:

> So... what are your comments/experiences?  How many eggs do you keep in
> your basket(s)?

As others have noted, I prefer to run one service (or interrelated set 
of services) per physical system instance.

Where the services support such things I also set up 
backup/secondaries/failovers, but there are still several services that 
either don't easily support having a backup or where the clients don't 
failover nicely.

For those services I prefer to build the best basket I can, so as to 
include as much redundancy as possible within the system itself. So, 
RAID1 or RAID5 for critical harddrives, redundant power supplies where 
possible, etc.  At the very least I'll try to keep a 'spare' system with 
the same hardware around and doing something non-critical as a potential 
source of immediate spare parts... on at least two occassions I've 
brought a more critical system back to life by transferring the hard 
drive into a 'hot spare' chassis and rebooting.

Since everything depends on our main NFS server so much, we spent the 
extra money to get a nice, highly internally redundant, fully 
service-contracted NetApp.  We also have ~6TB of what I call "scratch" 
filespace using less expensive (and less reliable) filesevers, which 
gives people room to store and analyze results but doesn't force 
everything to a screaming halt when something goes wrong.